Imaging Know-How Could Boost Microsoft Applications

Photogrammetry is technology that gleans information about objects by recording, measuring and interpreting images of them and patterns of electromagnetic or radiant energy and other phenomena.

Microsoft's purchase of Vexcel, Boulder, Colo., and Massive, New York, was announced Thursday as part of Microsoft's MSN Summit. Massive specializes in placing ads in video games. No terms were disclosed.

Vexcel's 2-D and 3-D imagery should bolster Microsoft's "Virtual Earth" business. In a statement, Stephen Lawler, general manager of Virtual Earth said "delivering best-of-breed local search and mapping offerings to consumers, businesses and government customers will depend on providing a rich, immersive and dynamic experience that mirrors what can be experienced in the real world." Virtual Earth competes with Google Maps.

Microsoft partners say the company is doing a lot of interesting enablement in several areas around mapping, geographical information systems (GIS) and related areas.

They cite this nascent Windows Live application. Local Live gives surfers both aerial and driver- (or pedestrian-) eye views of their on-screen locale. Such technology paired with a rich database of local businesses can provide an array of location-based services, partners said.

This service currently covers just downtown San Francisco and Seattle, but in the words of one partner, "this is dramatic technology. Microsoft has taken Google Maps and Keyhole and trumped them," he noted.

Last month, Paul Flessner, Microsoft's senior vice president of data and storage platforms, said one priority is expanding the current data store and database functions to handle not just text and tables, but images and sounds.

One stumbling block thus far has been a weakness in pattern- and image- recognition algorithms. That is an area to be addressed by third parties, and/or by Microsoft itself, he noted.

Right now if a user does a Google Image search, for example, the search hits on tagged keywords, "Britney Spears." In the future sophisticated databases and algorithms would enable a match on Ms. Spears' face. (Or whatever).

Data stores and databases will have to be redesigned or retrofitted to handle these content-rich data types and make them searchable on more than text tags, observers said. In Microsoft's case, the ability to handle that data and search it will come in the "Katmai" timeframe. Katmai is the next-gen SQL Server, expected by sources to debut in 2008.

A confluence of trends could enable an explosion of image-laden applications. Real estate searches could be done based on images of building types and locations; online shoppers looking for a specific article of clothing by style, color etc., could find it by dragging and dropping an image -- say a digital snapshot taken by camera phone -- to locate similar articles.

And then there are serious security and defense applications."There are absolutely military and industrial applications for that -- looking at slowly changing pictures may show construction of new capabilities of a 'party of interest.' Audio capture and comparison also has interesting capabilities. Combine audio search and you can now scan a recorded meeting for specific content points," says Lee Blackstone, CEO of Blackstone and Cullen, an Atlanta-based solution provider with database expertise. "Or scan 'syllabic content' for recorded calls of interest. It will be interesting to see how the meta-data will be created for this type of search. It would be interesting to have a way to compare or catalog electronic signals directly as opposed to having to run them thru DSP or Fourier Transforms first. I see a lot of apps for this....mainly way outside of the traditional IT organization."

Near-ubiquitous broadband and wireless links, an avalanche of camera-enabled devices, and cheap, available storage are all driving this trend, observers say.

Jeff Reed, CTO of Logicalis, a Bloomfield Hills, Mich.-based integrator, is bullish on the whole field.

"We're such a visual society and now companies like Amazon are digitizing everything. It's one thing to do an OCR [optical character recognition] on text-based stuff, it's entirely different [for the system] to recognize objects," he noted.

Insurance claims (both medical and property) could be expedited if pattern and image recognition technologies are honed, he said.

"Your doctor could take a picture of a mole on your back and match it against a known database of similar conditions. Once that is done, you might discover that nine out of 10 moles that resemble it were cancerous and proceed accordingly," he said.

"Or if you dent your car door, you photograph it and your insurance company matches that image against a catalog of similar damage done to the same vehicle," Reed noted.

Reed and others said database providers will need to adapt their wares to handle searching of these images but the technology is clearly on the way.

Bernie Spang, director of data servers for IBM Software's Information Management group says there are already real-world examples of image search and integration doing good work.

There was a case in New York City, where police were able to use imaging technology to get the bad guy, Spang said. "Police canvassed the crowd and someone saw a guy with a tattoo of the grim reaper with the word 'Shorty.' They input that and got back a series of images of tattoos which the officers showed the witness on the scene. They identified the tattoo graphics vs. known aliases of 'Shorty' and hit on a guy who fit the M.O."